Have you ever listened to an audio book and just thought : 'I wish someone else was reading this' or read a book while thinking : 'Why isn't that one actor reading this to me instead' ?
This project has for objective to satisfy this very first-world desire to get things just right.
The recent improvement in Deep Learning applications in style reproduction to images (http://arxiv.org/pdf/1508.06576v2.pdf) and voice generation (http://arxiv.org/pdf/1609.03499v1.pdf) have led me to believe that this kind of project is now within the realm of possibility.
I am a Computer Vision Scientist, working mostly with Deep Learning and specifically CNNs who wants to get really started on LSTMs and RNNs on the side.